Getting started

Install

For your agentCopy this and give it to the AI you're building with.

Set up cubek on this machine: install the CLI with `curl -fsSL https://cubek.dev/install | sh`, and bun (https://bun.sh) if it is missing.
Save the output of `cubek skill` as your cubek Skill and add the MCP server (Claude Code: `claude mcp add cubek -- cubek mcp`).
Run `cubek --help` to confirm, then tell me what you set up.

Your agent can do all of this from the prompt above. By hand: cubek is one CLI. It compiles your app locally, runs it with cubek dev, and deploys it. You need:

The cubek CLI: one binary, today for Linux on x86_64 and arm64 (glibc 2.28+); macOS builds are coming. The script checks its SHA-256 and installs it in ~/.local/bin (CUBEK_INSTALL_DIR to change it, CUBEK_VERSION to pin a version).
bun on the machine that deploys: it bundles browser code and runs cubek test.

curl -fsSL https://cubek.dev/install | sh
curl -fsSL https://bun.sh/install | bash
cubek --help

Give your agent the Skill

The Skill is a single page (about 4K tokens) that teaches a coding agent the whole platform. For Claude Code:

mkdir -p ~/.claude/skills/cubek
cubek skill > ~/.claude/skills/cubek/SKILL.md

For Cursor, save it as a project rule (.cursor/rules/cubek.mdc); for Codex and others, append it to the project's AGENTS.md.

Add the MCP server

The MCP server exposes the CLI as tools (check, deploy, logs, SQL...) plus sdk_types, the exact API as .d.ts:

claude mcp add cubek -- cubek mcp